CAD Archivium practice archive
ISO 286

Dimensional tolerances to ISO 286: the complete guide

What they are, how to read them, how to calculate them and how to choose them. The ISO 286 system from the ground up, with five worked examples, the deviation tables, the ISO 2768 general tolerances and the standards behind them.

Published 2026-08-13 17 min read

Every part of a flat-pack cabinet comes out of a different factory, often in a different country, and when you put them together they fit. That is not luck: it is the result of a precise language in which design tells production how far off it may be. That language is dimensional tolerancing.

The word comes from the Latin tolerare, to bear. Tolerances are the deviations we are willing to accept without losing the function of the part. They are not mistakes: they are deliberate design decisions, and they are how a functional requirement becomes something measurable and checkable.

This guide starts from the basics, works through the whole ISO 286 system, solves five worked examples step by step, and closes with the standards, the manufacturing processes and the mistakes not to make.

1. Why tolerances exist

One idea underpins everything: the exact part does not exist. Temperature swings, tool wear, machine vibration, the operator's skill — all of it introduces deviation. All you can do is define a band of acceptability inside which the actual size must fall for the part to do its job.

Take a shaft that must turn inside a hole. If the drawing says ⌀50 for both, and the workshop delivers a shaft at 50.004 and a hole at 49.998, the shaft does not fit. Nobody made a mistake: both parts are as close to 50 as anyone could reasonably ask. The mistake is in the drawing, which never said which side of 50 mattered.

The trade-off is always the same. Tight tolerances buy accuracy but cost more to make: better machines, more passes, stricter inspection, more scrap. Loose tolerances make production fast and cheap, at the risk of assemblies that do not work. A good designer does not ask for the highest accuracy available, but for the accuracy required.

Rule of thumb Cost does not rise linearly as you tighten: it steps, and the steps fall where the process has to change — turning to grinding, drilling to reaming. Tightening by one IT grade does not add ten per cent to the cost; it often doubles it. Always ask whether that accuracy is really needed for the function.

2. The minimum vocabulary

TermWhat it isExample
Nominal sizeThe size written on the drawing⌀50
Actual sizeWhat you measure on the finished part50.012
Upper deviationMaximum permitted minus nominalES (holes) / es (shafts)
Lower deviationMinimum permitted minus nominalEI (holes) / ei (shafts)
Tolerance zoneThe difference between the twoIT = ES − EI
Zero lineThe reference, that is the nominal50.000

One convention to learn straight away, because it is universal in technical drawing: capitals for holes, lowercase for shafts. If you read H7 you are looking at a hole; if you read h6, a shaft.

3. How tolerances appear on the drawing

3.1 Explicit deviations

The deviations written directly next to the dimension:

IndicationMeaning
⌀50 +0.025 / 0from 50.000 to 50.025
80 ±0.1from 79.900 to 80.100
25 +0.2 / −0.1from 24.900 to 25.200

This is the most direct and least ambiguous method, right for isolated dimensions or for anything that does not belong to a standard fit. Note the asymmetric case: +0.025/0 means the part may grow but must never fall below nominal. That is not a rounding convenience — it usually means something has to fit inside.

3.2 ISO tolerance classes

For mating dimensions the coded system is used: ⌀50H7 for a hole, ⌀50g6 for a shaft. The letter says where the tolerance zone sits relative to the zero line; the number says how wide it is.

3.3 General tolerances

Putting deviations on every single dimension would clutter the drawing until it became unreadable. For non-critical dimensions the general class is called up once, in the title block: General tolerances: ISO 2768-m. Every dimension without a tolerance of its own then follows that class automatically.

In daily practice: ISO classes for functional fits, explicit deviations for special dimensions, general tolerances for everything else.

4. The ISO 286 system

The system is defined by ISO 286-1, which sets out the basis, the terminology and the deviations, and by ISO 286-2, which holds the tables of values. It has two components: the tolerance grade and the position of the zone.

4.1 IT grades: how wide the zone is

There are eighteen, from IT01 to IT18. The lower the number, the tighter the tolerance. A grade is not a fixed number of millimetres: it widens with size, because holding ten micrometres on a 5 mm pin and on a 500 mm bore are not the same job.

Size range (mm)IT5IT6IT7IT8IT9IT10IT11IT12
up to 3461014254060100
over 3 to 6581218304875120
over 6 to 10691522365890150
over 10 to 1881118274370110180
over 18 to 3091321335284130210
over 30 to 501116253962100160250
over 50 to 801319304674120190300
over 80 to 1201522355487140220350
over 120 to 18018254063100160250400
over 180 to 25020294672115185290460
over 250 to 31523325281130210320520
over 315 to 40025365789140230360570
over 400 to 50027406397155250400630

Values in micrometres: 1 µm = 0.001 mm. Grades IT01 to IT4 are missing here because they belong to gauges and measuring instruments, not to general engineering parts; grades above IT12 cover rough processes and are listed in full in ISO 286-2.

The most expensive mistake in the subject The ranges read "over X up to and including Y". So ⌀30 belongs to the 18–30 range, not to 30–50, and ⌀50 belongs to 30–50, not to 50–80. Reading the wrong range produces numbers that are entirely plausible and entirely wrong: it is the one kind of error nobody catches when checking a drawing.

4.2 Positions: where the zone sits

The letters place the tolerance zone relative to the zero line.

LettersHoles (capitals)Shafts (lowercase)Effect
A to G / a to gabove zerobelow zeroclearance, from very large to minimal
H / hstarts at zero, risesstarts at zero, fallsthe reference of each system
JS / jsstraddling, symmetric (±IT/2)symmetric machining
J to N / j to nstraddlingstraddlingtransition fits
P to ZC / p to zcbelow zeroabove zerointerference, from light to very heavy

Note that holes and shafts move in opposite directions: for a shaft, the letters before h remove material and create clearance; for a hole, the letters before H add space and do the same thing. It is symmetric, but it confuses everyone the first time.

4.3 Fundamental deviations

For each position the standard gives one deviation, the one on the side of the zero line; the other follows from the IT grade. For letters a to h the tabulated value is the upper deviation es, and the lower one follows by subtracting the IT. From k onwards the tabulated value is the lower deviation ei, and the upper one follows by adding the IT.

Range (mm)d (es)e (es)f (es)g (es)h (es)k (ei)m (ei)n (ei)p (ei)s (ei)
over 6 to 10−40−25−13−50+1+6+10+15+23
over 10 to 18−50−32−16−60+1+7+12+18+28
over 18 to 30−65−40−20−70+2+8+15+22+35
over 30 to 50−80−50−25−90+2+9+17+26+43

Values for shafts, in micrometres. For holes the same figures apply with the sign reversed and the matching capital, except for positions J, K, M and N at the finer grades, where ISO 286-2 introduces a correction: those must be read from the table, not worked out in your head. Above 50 mm some positions subdivide the size ranges further — at 65 and at 100 mm, for instance — so beyond that point you consult the standard rather than extrapolate.

4.4 Hole basis or shaft basis

On a hole basis the hole stays fixed at H and the fit is changed by moving the shaft: H7/g6, H7/k6, H7/p6. On a shaft basis the shaft is fixed at h and the hole moves: F8/h7, K7/h6, P7/h6.

The first is used in the overwhelming majority of cases, and the reason is economic. A hole is produced with a reamer or a broach of a definite size, and changing its tolerance means buying a new tool; a shaft can be turned to any diameter with the tool you already have. Shaft basis earns its place when one through shaft has to mate with several different holes: then it pays to hold the shaft still and move the holes.

5. The three kinds of fit

Mating a hole with a shaft gives one of three situations.

FitTypeApplicationAssembly
H11/c11very large clearancefabrications, farm machineryby hand
H9/d9large clearancelinkages, slow jointsby hand
H8/f7running clearanceplain bearings, slow shaftsby hand
H7/g6close runningguides, pistons, cylindersby hand
H7/h6minimal clearanceremovable locationby hand
H7/js6transitionfrequent dismantlingby hand or mallet
H7/k6transitionbearing seats, pinsmallet
H7/n6light interferencefixed locationmallet or press
H7/p6medium interferencepins, bushespress
H7/s6heavy interferenceshrink fitspress or heat
H7/u6very heavy interferencegear rings, flywheelsheat

6. Five worked examples

Example 1 — Clearance fit: ⌀50 H7/g6

A piston sliding in a cylinder: enough clearance is needed for the oil film. Diameter 50, so the range is over 30 to 50.

  1. Hole H7. From the table, IT7 = 25 µm. Position H forces EI = 0, so ES = +0.025 mm. Hole from 50.000 to 50.025 mm.
  2. Shaft g6. IT6 = 16 µm; position g gives es = −9 µm, so ei = −9 − 16 = −25 µm. Shaft from 49.975 to 49.991 mm.
  3. Minimum clearance = smallest hole − largest shaft = 50.000 − 49.991 = 0.009 mm.
    Maximum clearance = largest hole − smallest shaft = 50.025 − 49.975 = 0.050 mm.

The clearance stays positive throughout, between nine and fifty micrometres: every part made to that drawing slides, none seizes and none rattles.

Example 2 — Transition fit: ⌀25 H7/k6

A bearing seat: accurate location, still removable. Diameter 25, range over 18 to 30.

  1. Hole H7. IT7 = 21 µm, EI = 0, ES = +0.021 mm. Hole from 25.000 to 25.021 mm.
  2. Shaft k6. IT6 = 13 µm; k gives ei = +2 µm, so es = +2 + 13 = +15 µm. Shaft from 25.002 to 25.015 mm.
  3. Worst case one way: smallest hole with largest shaft, 0.015 mm of interference.
    Worst case the other way: largest hole with smallest shaft, 0.019 mm of clearance.

You do not know in advance which of the two you will get — hence the name. The bearing will be located accurately and will come off with a puller.

Example 3 — Interference fit: ⌀40 H7/s6

A gear ring shrunk onto a shaft: a permanent joint. Diameter 40, range over 30 to 50.

  1. Hole H7. IT7 = 25 µm. Hole from 40.000 to 40.025 mm.
  2. Shaft s6. IT6 = 16 µm; s gives ei = +43 µm, so es = +43 + 16 = +59 µm. Shaft from 40.043 to 40.059 mm.
  3. Minimum interference = smallest shaft − largest hole = 40.043 − 40.025 = 0.018 mm.
    Maximum interference = largest shaft − smallest hole = 40.059 − 40.000 = 0.059 mm.

The shaft is always larger than the hole: it needs a press, or the ring is heated to expand it, slipped on and left to cool. Mind the order of the terms: minimum interference comes from the combination most favourable to assembly — small shaft in large hole — not from the larger absolute number.

Example 4 — Explicit deviations: 65 +0.04 / −0.02

A functional dimension that does not belong to a standard fit, a shoulder for instance.

  1. Maximum size = 65 + 0.04 = 65.04 mm; minimum = 65 − 0.02 = 64.98 mm.
  2. Tolerance zone = ES − EI = 0.04 − (−0.02) = 0.06 mm, that is 60 µm.

In the 50–80 range that value falls between IT8 (46 µm) and IT9 (74 µm), so the dimension is looser than IT8 and tighter than IT9. It is straightforward turning work.

Example 5 — Shaft basis: ⌀30 F8/h7

One through shaft that has to mate with several different holes. Diameter 30 — careful, it falls in the over 18 to 30 range.

  1. Shaft h7. IT7 = 21 µm; h gives es = 0, so ei = −0.021 mm. Shaft from 29.979 to 30.000 mm.
  2. Hole F8. IT8 = 33 µm; position F gives EI = +20 µm, so ES = +20 + 33 = +53 µm. Hole from 30.020 to 30.053 mm.
  3. Minimum clearance = 30.020 − 30.000 = 0.020 mm.
    Maximum clearance = 30.053 − 29.979 = 0.074 mm.

The shaft is the reference and the hole provides the clearance. To mate the same shaft with another component you simply change the hole position: G8, H8, and so on.

7. General tolerances to ISO 2768

For non-critical dimensions ISO 2768-1 defines four classes of accuracy, called up once in the title block.

7.1 Linear dimensions

Range (mm)f (fine)m (medium)c (coarse)v (very coarse)
0.5 to 3±0.05±0.1±0.2
over 3 to 6±0.05±0.1±0.3±0.5
over 6 to 30±0.1±0.2±0.5±1
over 30 to 120±0.15±0.3±0.8±1.5
over 120 to 400±0.2±0.5±1.2±2.5
over 400 to 1000±0.3±0.8±2±4
over 1000 to 2000±0.5±1.2±3±6

7.2 Angular dimensions

Shorter side (mm)fmcv
up to 10±1°±1°±1°30′±3°
over 10 to 50±30′±30′±1°±2°
over 50 to 120±20′±20′±30′±1°
over 120 to 400±10′±10′±15′±30′
over 400±5′±5′±10′±20′

7.3 General geometrical tolerances

ISO 2768-2 covers geometry with classes H, K and L. Straightness and flatness depend on the length of the feature; they are not a single value:

Nominal length (mm)HKL
up to 100.020.050.1
over 10 to 300.050.10.2
over 30 to 1000.10.20.4
over 100 to 3000.20.40.8
over 300 to 10000.30.61.2

Perpendicularity and symmetry also depend on the length of the side; circular run-out, by contrast, is a single value per class: 0.1 for H, 0.2 for K, 0.5 for L.

Two things summaries almost always get wrong General circularity has no table of its own: it equals the size tolerance of the diameter, and in any case it cannot exceed the circular run-out value. General cylindricity is not defined by ISO 2768-2: it results from circularity combined with straightness and the parallelism of the generating lines. If you find a table giving a cylindricity value attributed to this standard, that table is wrong.

Choosing the class: f for precision engineering, m for general engineering — the right answer in the great majority of cases — c for fabrication and sheet metal, v for castings and rough parts. A drawing marked ISO 2768-mK has covered linear and geometric in six characters.

Read that in reverse and you have the discipline the standard is really teaching: if a dimension carries its own tolerance, it is because something depends on it. A drawing where every dimension is toleranced individually is a drawing whose author had not decided what mattered.

8. The standards

StandardSubjectWhat is in it
ISO 286-1System of limits and fitsBasis, terminology, IT grades, positions
ISO 286-2Tables of deviationsValues for holes and shafts up to IT18
ISO 2768-1General dimensional tolerancesClasses f, m, c, v — linear and angular
ISO 2768-2General geometrical tolerancesClasses H, K, L
ISO 1101Geometrical tolerancingForm, orientation, location, run-out
ISO 8015Fundamental GPS principlesThe principle of independency
ISO 14405-1Definition of sizeThe Ⓔ symbol, measurement criteria
ISO 129-1DimensioningRules for indication on drawings
ASME Y14.5Dimensioning and tolerancing (USA)Rule #1, American GD&T
ISO versus ASME: a difference that changes the part Under ISO the default is the principle of independency: size and geometry are separate requirements, and a shaft can be within its size tolerance and still be bent. Under ASME the default is the envelope requirement (Rule #1): form is automatically controlled by size at maximum material condition. The same drawing, read under the wrong standard, describes two different parts. Under ISO the envelope requirement can be called up explicitly with the Ⓔ symbol.

9. What accuracy each process gives

Every process reaches a characteristic band of IT grades. Knowing it is the only way to specify tolerances that can actually be met.

ProcessIT gradesTypical applications
Grinding, lapping, honingIT4 to IT6precision shafts, bearing seats, gauges
Precision reaming and boringIT6 to IT7bearing bores, hydraulic cylinders
Fine turning, CNC millingIT7 to IT8shafts, flanges, precision parts
Ordinary turningIT8 to IT9pins, screws, general parts
Ordinary milling, drilling with a reamerIT9 to IT11clearance holes, overall dimensions
Plain drillingIT11 to IT12non-functional through holes
Laser, plasma and flame cuttingIT12 to IT14sheet profiles, structural shapes
Bending, deep drawing, stampingIT13 to IT16sheet metal work, fabrications
Die casting, permanent mould castingIT12 to IT14high volumes, complex shapes
Sand castingIT14 to IT16rough parts, machine bodies

Hence the question worth asking before writing a tolerance: which process will the workshop use? Asking for IT5 on a sand casting is asking for the impossible; writing IT12 on a precision shaft means wasting the grinding operation that will be carried out anyway.

10. Common mistakes

The ten per cent rule The measuring instrument should have an uncertainty of roughly one tenth of the tolerance being checked. For a tolerance of 0.02 mm you need an instrument good to about 0.002 mm: a micrometer or a dial indicator reading in microns, not a vernier caliper. For IT8–IT9 a micrometer and a precision caliper are enough; for IT10–IT11 a standard caliper will do.

11. In short

Tolerances are rather like traffic rules: they look as though they restrict freedom, and in fact they are what makes it possible. Without them every part would go its own way and no assembly would stand up.

12. Putting it on a sheet

Reading only takes you so far. Take a sheet from the archive with a shaft and a housing, decide for yourself which dimensions are functional, choose fits for those and a general class for everything else, and write it on the drawing. Then change your mind about one fit and see how far the consequences travel.

That last step is the one that teaches. A tolerance is never a property of a single feature: it is a statement about a relationship, and relationships are where drawings get interesting.

The next article in the series will take on geometrical tolerancing to ISO 1101: what happens when size alone is not enough to describe what a part has to do.

Go to the archive All articles